【杨勇】小论文MATLAB实验程序主要部分代码--语音增强--stsa_weuclid1

function stsa_weuclid1(filename,outfile,p)

if nargin<3 fprintf('Usage: stsa_weuclid inFile outFile.wav p \n\n'); return; end

fp=fopen(filename,'r'); if fp<=0, error('Unable to open input file'); end;

% ------------ get the file extension, and then read the whole file ----------- % ind1=find(filename == '.'); if length(ind1)>1, ind=ind1(length(ind1)); else, ind=ind1; end; ext = lower(filename(ind+1:length(filename)));

[HDRSIZE, Srate, bpsa, ftype] = gethdr(fp,ext); x=fread(fp,inf,'short'); fclose(fp);

%x=x/32768;

% =============== Initialize variables =============== %

len=floor(20Srate/1000); % Frame size in samples if rem(len,2)==1, len=len+1; end; PERC=50; % window overlap in percent of frame size len1=floor(lenPERC/100); len2=len-len1;

win=hanning(len); %tukey(len,PERC); % define window U=norm(win);

% Noise magnitude calculations - assuming that the first 6 frames is noise/silence % nFFT=len; nFFT2=len/2; noise_mean=zeros(nFFT,1); j=1; for k=1:5 noise_mean=noise_mean+abs(fft(win.*x(j:j+len-1),nFFT)/U); j=j+len; end noise_mu=noise_mean/5; noise_mu2=noise_mu.^2;

%--- allocate memory and initialize various variables

扫描二维码关注公众号，回复： 410706 查看本文章

k=1; img=sqrt(-1); x_old=zeros(len1,1); Nframes=floor(length(x)/len2)-1; xfinal=zeros(Nframes*len2,1);

fr=Srate/nFFT; fIndx=1:floor(4000/fr); freqs=0:fr:(4000-fr);

%=============================== Start Processing ======================================================= % k=1; aa=0.98; c=sqrt(pi)/2; ksi_ar=zeros(Nframes,1); gam_ar=ksi_ar; C2=gamma(0.5);

%p=-1.0; CC=gamma(0.6)/gamma(0.1); STSA=0; for n=1:Nframes

insign=win.*x(k:k+len-1);

%--- Take fourier transform of frame

spec=fft(insign,nFFT)/U;
sig=abs(spec); % compute the magnitude sig2=sig.^2;

   gammak=min(sig2./noise_mu2,40);  % post SNR
   if n==1
       ksi=aa+(1-aa)*max(gammak-1,0);
   else

       ksi=aa*Xk_prev./noise_mu2 + (1-aa)*max(gammak-1,0);     % a priori SNR   
      %ksi=max(gammak-1,0);
   end

   vv=0.6;
   vk=ksi.*gammak./(vv+ksi);

  %----- weighted Euclidean distance ------------------------
    numer=CC*sqrt(vk).*confhyperg(0.6,1,vk,100);
    denom=gammak.*confhyperg(0.1,1,vk,100);
    hw=numer./denom;

    %hw2=CC*sqrt(vk)./(gammak.*exp(-vk/2).*besseli(0,vk/2));  % if p=-1

    %       
   % ---- for the cosh measure

% numer = sqrt(gamma(1.5)*vk.*confhyperg(-0.5,1,-vk,40)); % denom = gammak.*sqrt(gamma(0.5)*confhyperg(0.5,1,-vk,40)); % hw=numer./denom;

   % --- for the weighted cosh measure
   %CC2=gamma((p+3)/2)/gamma((p+1)/2);
   %numer=CC2*sqrt(vk).*sqrt(confhyperg(-(p+1)/2,1,-vk,40));
   %denom=gammak.*sqrt(confhyperg(-(p-1)/2,1,-vk,40));
   %hw=numer./denom;

   sig=sig.*hw;
   Xk_prev=sig.^2;

   xi_w= ifft( hw .* spec); 
   xi_w= real( xi_w)*U;

% --- Overlap and add ---------------
%
xfinal(k:k+ len2-1)= x_old+ xi_w(1:len1);
x_old= xi_w(len1+ 1: len);

k=k+len2; end %========================================================================================

if max(abs(xfinal))>32768 xfinal=xfinal*24000/max(abs(xfinal)); fprintf('Max amplitude exceeded 32768 for file %s\n',filename);
end xfinal=xfinal/32768;

wavwrite(xfinal,Srate,16,outfile);

return;

% -------- plot SNR --------------- % timex=0:1000/Srate:Nframes*10; subplot(2,1,1),h=plot(timex,x(1:length(timex))); set(gca,'Box','off','Xlim',[0 timex(length(timex))],'FontSize',12); ylabel('Amplitude');

time2=0:10:(Nframes-1)10; subplot(2,1,2), plot(time2,10log10(hgain));

h=plot(time2,10log10(ksi_ar),'b',time2,10log10(gam_ar),'r:');

legend('\xi_k','\gamma_k-1'); set(h,'Linewidth',1.5); set(gca,'FontSize',12,'Box','off','Xlim',[0 time2(length(time2))],'Ylim',[-25 max(10*log10(gam_ar))]); xlabel('Time (ms)'); ylabel(' SNR (dB)');

return;

subplot(3,1,3), %h=plot(time2,20*log10(hgain));

h=plot(20*log10(hgain)); %set(gca,'Box','off','Xlim',[0 timex(length(timex))],'FontSize',12); ylabel('Amplitude');

【杨勇】小论文MATLAB实验程序主要部分代码--语音增强--stsa_weuclid1

猜你喜欢